/*
* do_bdr_register()
*
* As each BDR node is its own primary, registering a BDR node
* will create the repmgr metadata schema if necessary.
*/
void
do_bdr_register(void)
{
PGconn *conn = NULL;
BdrNodeInfoList bdr_nodes = T_BDR_NODE_INFO_LIST_INITIALIZER;
ExtensionStatus extension_status = REPMGR_UNKNOWN;
t_node_info node_info = T_NODE_INFO_INITIALIZER;
RecordStatus record_status = RECORD_NOT_FOUND;
PQExpBufferData event_details;
bool success = true;
char *dbname = NULL;
/* sanity-check configuration for BDR-compatability */
if (config_file_options.replication_type != REPLICATION_TYPE_BDR)
{
log_error(_("cannot run BDR REGISTER on a non-BDR node"));
exit(ERR_BAD_CONFIG);
}
dbname = pg_malloc0(MAXLEN);
if (dbname == NULL)
{
log_error(_("unable to allocate memory; terminating."));
exit(ERR_OUT_OF_MEMORY);
}
/* store the database name for future reference */
get_conninfo_value(config_file_options.conninfo, "dbname", dbname);
conn = establish_db_connection(config_file_options.conninfo, true);
if (!is_bdr_db(conn, NULL))
{
log_error(_("database \"%s\" is not BDR-enabled"), dbname);
log_hint(_("when using repmgr with BDR, the repmgr schema must be stored in the BDR database"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
/* Check that there are at most 2 BDR nodes */
get_all_bdr_node_records(conn, &bdr_nodes);
if (bdr_nodes.node_count == 0)
{
log_error(_("database \"%s\" is BDR-enabled but no BDR nodes were found"), dbname);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
/* BDR 2 implementation is for 2 nodes only */
if (get_bdr_version_num() < 3 && bdr_nodes.node_count > 2)
{
log_error(_("repmgr can only support BDR 2.x clusters with 2 nodes"));
log_detail(_("this BDR cluster has %i nodes"), bdr_nodes.node_count);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
if (get_bdr_version_num() > 2)
{
log_error(_("\"repmgr bdr register\" is for BDR 2.x only"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
/* check for a matching BDR node */
{
PQExpBufferData bdr_local_node_name;
bool node_match = false;
initPQExpBuffer(&bdr_local_node_name);
node_match = bdr_node_name_matches(conn, config_file_options.node_name, &bdr_local_node_name);
if (node_match == false)
{
if (strlen(bdr_local_node_name.data))
{
log_error(_("local node BDR node name is \"%s\", expected: \"%s\""),
bdr_local_node_name.data,
config_file_options.node_name);
log_hint(_("\"node_name\" in repmgr.conf must match \"node_name\" in bdr.bdr_nodes"));
}
else
{
log_error(_("local node does not report BDR node name"));
log_hint(_("ensure this is an active BDR node"));
}
PQfinish(conn);
pfree(dbname);
termPQExpBuffer(&bdr_local_node_name);
exit(ERR_BAD_CONFIG);
}
termPQExpBuffer(&bdr_local_node_name);
}
/* check whether repmgr extension exists, and there are no non-BDR nodes registered */
extension_status = get_repmgr_extension_status(conn, NULL);
if (extension_status == REPMGR_UNKNOWN)
{
log_error(_("unable to determine status of \"repmgr\" extension in database \"%s\""),
dbname);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
if (extension_status == REPMGR_UNAVAILABLE)
{
log_error(_("\"repmgr\" extension is not available"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
if (extension_status == REPMGR_INSTALLED)
{
if (!is_bdr_repmgr(conn))
{
log_error(_("repmgr metadatabase contains records for non-BDR nodes"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
}
else
{
log_debug("creating repmgr extension in database \"%s\"", dbname);
begin_transaction(conn);
if (!create_repmgr_extension(conn))
{
log_error(_("unable to create repmgr extension - see preceding error message(s); aborting"));
rollback_transaction(conn);
pfree(dbname);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
commit_transaction(conn);
}
pfree(dbname);
if (bdr_node_has_repmgr_set(conn, config_file_options.node_name) == false)
{
log_debug("bdr_node_has_repmgr_set() = false");
bdr_node_set_repmgr_set(conn, config_file_options.node_name);
}
/*
* before adding the extension tables to the replication set, if any other
* BDR nodes exist, populate repmgr.nodes with a copy of existing entries
*
* currently we won't copy the contents of any other tables
*
*/
{
NodeInfoList local_node_records = T_NODE_INFO_LIST_INITIALIZER;
(void) get_all_node_records(conn, &local_node_records);
if (local_node_records.node_count == 0)
{
BdrNodeInfoList bdr_nodes = T_BDR_NODE_INFO_LIST_INITIALIZER;
BdrNodeInfoListCell *bdr_cell = NULL;
get_all_bdr_node_records(conn, &bdr_nodes);
if (bdr_nodes.node_count == 0)
{
log_error(_("unable to retrieve any BDR node records"));
log_detail("%s", PQerrorMessage(conn));
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
for (bdr_cell = bdr_nodes.head; bdr_cell; bdr_cell = bdr_cell->next)
{
PGconn *bdr_node_conn = NULL;
NodeInfoList existing_nodes = T_NODE_INFO_LIST_INITIALIZER;
NodeInfoListCell *cell = NULL;
ExtensionStatus other_node_extension_status = REPMGR_UNKNOWN;
/* skip the local node */
if (strncmp(node_info.node_name, bdr_cell->node_info->node_name, sizeof(node_info.node_name)) == 0)
{
continue;
}
log_debug("connecting to BDR node \"%s\" (conninfo: \"%s\")",
bdr_cell->node_info->node_name,
bdr_cell->node_info->node_local_dsn);
bdr_node_conn = establish_db_connection_quiet(bdr_cell->node_info->node_local_dsn);
if (PQstatus(bdr_node_conn) != CONNECTION_OK)
{
continue;
}
/* check repmgr schema exists, skip if not */
other_node_extension_status = get_repmgr_extension_status(bdr_node_conn, NULL);
if (other_node_extension_status != REPMGR_INSTALLED)
{
continue;
}
(void) get_all_node_records(bdr_node_conn, &existing_nodes);
for (cell = existing_nodes.head; cell; cell = cell->next)
{
log_debug("creating record for node \"%s\" (ID: %i)",
cell->node_info->node_name, cell->node_info->node_id);
create_node_record(conn, "bdr register", cell->node_info);
}
PQfinish(bdr_node_conn);
break;
}
}
}
/* Add the repmgr extension tables to a replication set */
if (get_bdr_version_num() < 3)
{
add_extension_tables_to_bdr_replication_set(conn);
}
else
{
/* this is the only table we need to replicate */
char *replication_set = get_default_bdr_replication_set(conn);
/*
* this probably won't happen, but we need to be sure we're using
* the replication set metadata correctly...
*/
if (conn == NULL)
{
log_error(_("unable to retrieve default BDR replication set"));
log_hint(_("see preceding messages"));
log_debug("check query in get_default_bdr_replication_set()");
exit(ERR_BAD_CONFIG);
}
if (is_table_in_bdr_replication_set(conn, "nodes", replication_set) == false)
{
add_table_to_bdr_replication_set(conn, "nodes", replication_set);
}
pfree(replication_set);
}
initPQExpBuffer(&event_details);
begin_transaction(conn);
/*
* we'll check if a record exists (even if the schema was just created),
* as there's a faint chance of a race condition
*/
record_status = get_node_record(conn, config_file_options.node_id, &node_info);
/* Update internal node record */
node_info.type = BDR;
node_info.node_id = config_file_options.node_id;
node_info.upstream_node_id = NO_UPSTREAM_NODE;
node_info.active = true;
node_info.priority = config_file_options.priority;
strncpy(node_info.node_name, config_file_options.node_name, sizeof(node_info.node_name));
strncpy(node_info.location, config_file_options.location, sizeof(node_info.location));
strncpy(node_info.conninfo, config_file_options.conninfo, sizeof(node_info.conninfo));
if (record_status == RECORD_FOUND)
{
bool node_updated = false;
/*
* At this point we will have established there are no non-BDR
* records, so no need to verify the node type
*/
if (!runtime_options.force)
{
log_error(_("this node is already registered"));
log_hint(_("use -F/--force to overwrite the existing node record"));
rollback_transaction(conn);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
/*
* don't permit changing the node name - this must match the BDR node
* name set when the node was registered.
*/
if (strncmp(node_info.node_name, config_file_options.node_name, sizeof(node_info.node_name)) != 0)
{
log_error(_("a record for node %i is already registered with node_name \"%s\""),
config_file_options.node_id, node_info.node_name);
log_hint(_("node_name configured in repmgr.conf is \"%s\""), config_file_options.node_name);
rollback_transaction(conn);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
node_updated = update_node_record(conn, "bdr register", &node_info);
if (node_updated == true)
{
appendPQExpBuffer(&event_details, _("node record updated for node \"%s\" (%i)"),
config_file_options.node_name, config_file_options.node_id);
log_verbose(LOG_NOTICE, "%s", event_details.data);
}
else
{
success = false;
}
}
else
{
/* create new node record */
bool node_created = create_node_record(conn, "bdr register", &node_info);
if (node_created == true)
{
appendPQExpBuffer(&event_details,
_("node record created for node \"%s\" (ID: %i)"),
config_file_options.node_name, config_file_options.node_id);
log_notice("%s", event_details.data);
}
else
{
success = false;
}
}
if (success == false)
{
rollback_transaction(conn);
PQfinish(conn);
exit(ERR_DB_QUERY);
}
commit_transaction(conn);
/* Log the event */
create_event_notification(
conn,
&config_file_options,
config_file_options.node_id,
"bdr_register",
true,
event_details.data);
termPQExpBuffer(&event_details);
PQfinish(conn);
log_notice(_("BDR node %i registered (conninfo: %s)"),
config_file_options.node_id, config_file_options.conninfo);
return;
}
/*
* _build_single_nodeline_info - From the slurm.conf reader, build table,
* and set values
* RET 0 if no error, error code otherwise
* Note: Operates on common variables
* default_node_record - default node configuration values
*/
static int _build_single_nodeline_info(slurm_conf_node_t *node_ptr,
struct config_record *config_ptr)
{
int error_code = SLURM_SUCCESS;
struct node_record *node_rec = NULL;
hostlist_t address_list = NULL;
hostlist_t alias_list = NULL;
hostlist_t hostname_list = NULL;
hostlist_t port_list = NULL;
char *address = NULL;
char *alias = NULL;
char *hostname = NULL;
char *port_str = NULL;
int state_val = NODE_STATE_UNKNOWN;
int address_count, alias_count, hostname_count, port_count;
uint16_t port = 0;
if (node_ptr->state != NULL) {
state_val = state_str2int(node_ptr->state, node_ptr->nodenames);
if (state_val == NO_VAL)
goto cleanup;
}
if ((address_list = hostlist_create(node_ptr->addresses)) == NULL) {
fatal("Unable to create NodeAddr list from %s",
node_ptr->addresses);
error_code = errno;
goto cleanup;
}
if ((alias_list = hostlist_create(node_ptr->nodenames)) == NULL) {
fatal("Unable to create NodeName list from %s",
node_ptr->nodenames);
error_code = errno;
goto cleanup;
}
if ((hostname_list = hostlist_create(node_ptr->hostnames)) == NULL) {
fatal("Unable to create NodeHostname list from %s",
node_ptr->hostnames);
error_code = errno;
goto cleanup;
}
if (node_ptr->port_str && node_ptr->port_str[0] &&
(node_ptr->port_str[0] != '[') &&
(strchr(node_ptr->port_str, '-') ||
strchr(node_ptr->port_str, ','))) {
xstrfmtcat(port_str, "[%s]", node_ptr->port_str);
port_list = hostlist_create(port_str);
xfree(port_str);
} else {
port_list = hostlist_create(node_ptr->port_str);
}
if (port_list == NULL) {
error("Unable to create Port list from %s",
node_ptr->port_str);
error_code = errno;
goto cleanup;
}
/* some sanity checks */
address_count = hostlist_count(address_list);
alias_count = hostlist_count(alias_list);
hostname_count = hostlist_count(hostname_list);
port_count = hostlist_count(port_list);
#ifdef HAVE_FRONT_END
if ((hostname_count != alias_count) && (hostname_count != 1)) {
error("NodeHostname count must equal that of NodeName "
"records of there must be no more than one");
goto cleanup;
}
if ((address_count != alias_count) && (address_count != 1)) {
error("NodeAddr count must equal that of NodeName "
"records of there must be no more than one");
goto cleanup;
}
#else
#ifdef MULTIPLE_SLURMD
if ((address_count != alias_count) && (address_count != 1)) {
error("NodeAddr count must equal that of NodeName "
"records of there must be no more than one");
goto cleanup;
}
#else
if (address_count < alias_count) {
error("At least as many NodeAddr are required as NodeName");
goto cleanup;
}
if (hostname_count < alias_count) {
error("At least as many NodeHostname are required "
"as NodeName");
goto cleanup;
}
#endif /* MULTIPLE_SLURMD */
#endif /* HAVE_FRONT_END */
if ((port_count != alias_count) && (port_count > 1)) {
error("Port count must equal that of NodeName "
"records or there must be no more than one");
goto cleanup;
}
/* now build the individual node structures */
while ((alias = hostlist_shift(alias_list))) {
if (address_count > 0) {
address_count--;
if (address)
free(address);
address = hostlist_shift(address_list);
}
if (hostname_count > 0) {
hostname_count--;
if (hostname)
free(hostname);
hostname = hostlist_shift(hostname_list);
}
if (port_count > 0) {
int port_int;
port_count--;
if (port_str)
free(port_str);
port_str = hostlist_shift(port_list);
port_int = atoi(port_str);
if ((port_int <= 0) || (port_int > 0xffff))
fatal("Invalid Port %s", node_ptr->port_str);
port = port_int;
}
/* find_node_record locks this to get the
* alias so we need to unlock */
node_rec = find_node_record(alias);
if (node_rec == NULL) {
node_rec = create_node_record(config_ptr, alias);
if ((state_val != NO_VAL) &&
(state_val != NODE_STATE_UNKNOWN))
node_rec->node_state = state_val;
node_rec->last_response = (time_t) 0;
node_rec->comm_name = xstrdup(address);
node_rec->node_hostname = xstrdup(hostname);
node_rec->port = port;
node_rec->weight = node_ptr->weight;
node_rec->features = xstrdup(node_ptr->feature);
node_rec->reason = xstrdup(node_ptr->reason);
} else {
/* FIXME - maybe should be fatal? */
error("Reconfiguration for node %s, ignoring!", alias);
}
free(alias);
}
/* free allocated storage */
cleanup:
if (address)
free(address);
if (hostname)
free(hostname);
if (port_str)
free(port_str);
if (address_list)
hostlist_destroy(address_list);
if (alias_list)
hostlist_destroy(alias_list);
if (hostname_list)
hostlist_destroy(hostname_list);
if (port_list)
hostlist_destroy(port_list);
return error_code;
}
/*
* copy_configuration()
*
* Copy records in master's `repl_nodes` table to witness database
*
* This is used by `repmgr` when setting up the witness database, and
* `repmgrd` after a failover event occurs
*/
bool
copy_configuration(PGconn *masterconn, PGconn *witnessconn, char *cluster_name)
{
char sqlquery[MAXLEN];
PGresult *res;
int i;
sqlquery_snprintf(sqlquery, "TRUNCATE TABLE %s.repl_nodes", get_repmgr_schema_quoted(witnessconn));
log_debug("copy_configuration: %s\n", sqlquery);
res = PQexec(witnessconn, sqlquery);
if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
{
fprintf(stderr, "Cannot clean node details in the witness, %s\n",
PQerrorMessage(witnessconn));
return false;
}
sqlquery_snprintf(sqlquery,
"SELECT id, type, upstream_node_id, name, conninfo, priority, slot_name FROM %s.repl_nodes",
get_repmgr_schema_quoted(masterconn));
res = PQexec(masterconn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
fprintf(stderr, "Can't get configuration from master: %s\n",
PQerrorMessage(masterconn));
PQclear(res);
return false;
}
for (i = 0; i < PQntuples(res); i++)
{
bool node_record_created;
char *witness = PQgetvalue(res, i, 4);
log_debug(_("copy_configuration(): %s\n"), witness);
node_record_created = create_node_record(witnessconn,
"copy_configuration",
atoi(PQgetvalue(res, i, 0)),
PQgetvalue(res, i, 1),
strlen(PQgetvalue(res, i, 2))
? atoi(PQgetvalue(res, i, 2))
: NO_UPSTREAM_NODE,
cluster_name,
PQgetvalue(res, i, 3),
PQgetvalue(res, i, 4),
atoi(PQgetvalue(res, i, 5)),
strlen(PQgetvalue(res, i, 6))
? PQgetvalue(res, i, 6)
: NULL
);
if (node_record_created == false)
{
fprintf(stderr, "Unable to copy node record to witness database: %s\n",
PQerrorMessage(witnessconn));
return false;
}
}
PQclear(res);
return true;
}